Re: [PATCH 1/1] drm/amdgpu: free hw_vm_fence when fail in amdgpu_job_alloc

[Christian König]

On 1/14/26 11:17, Jiqian Chen wrote:
> If drm_sched_job_init fails, hw_vm_fence is not freed currently,
> then cause memory leak.
> 
> Signed-off-by: Jiqian Chen <jiqian.c...@amd.com>
> ---
>  drivers/gpu/drm/amd/amdgpu/amdgpu_job.c | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c 
> b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
> index 7f5d01164897..a5d3b35353da 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c
> @@ -216,11 +216,11 @@ int amdgpu_job_alloc(struct amdgpu_device *adev, struct 
> amdgpu_vm *vm,
>       (*job)->generation = amdgpu_vm_generation(adev, vm);
>       (*job)->vm_pd_addr = AMDGPU_BO_INVALID_OFFSET;
>  
> -     if (!entity)
> +     if (!entity ||
> +             !drm_sched_job_init(&(*job)->base, entity, 1, owner, 
> drm_client_id))
>               return 0;

That is not even remotely correct.

You need to assign the return code and check that instead, otherwise you return 
success when a failure occurred and just crash.

Regards,
Christian.

>  
> -     return drm_sched_job_init(&(*job)->base, entity, 1, owner,
> -                               drm_client_id);
> +     kfree((*job)->hw_vm_fence);
>  
>  err_fence:
>       kfree((*job)->hw_fence);